Early Learning Curve of Robotic-Assisted Radical Prostatectomy using CUSUM Analysis: Experience from a Tertiary Care Centre
摘要
Robotic-assisted radical prostatectomy (RARP) is a technically demanding procedure with a substantial learning curve, particularly during initial programme implementation. Objective assessment of performance evolution during early adoption is essential for quality assurance and benchmarking. This study evaluates the early institutional learning curve of RARP using cumulative sum (CUSUM) analysis at a tertiary care centre. A retrospective analysis of 125 consecutive RARP procedures performed between October 2023 and October 2025 by two consultant urologists at a single institution was conducted. The study was approved by the Institutional Ethics Committee of Deenanath Mangeshkar Hospital and Research Centre. CUSUM analysis was applied to console time and positive surgical margin (PSM) status. The phase division was determined by the CUSUM inflection points: the PSM CUSUM peaked at case 61 and the console time CUSUM reached its deepest trough at case 62; based on this convergent evidence, cases were divided at case 62. Stage-stratified PSM analysis (pT2 vs. pT3) was performed. The mean console time was 218.4 ± 29.1 min. The overall PSM rate was 18.4% (23/125), decreasing from 29.0% in the early phase to 7.9% in the later phase (p = 0.003). Stage-stratified analysis showed that this improvement was driven by pT3 disease (40.0% to 8.5%; p < 0.001). Phase-stratified functional analysis at 6 months demonstrated a trend toward improved potency recovery (SHIM ≥ 17: 37.1% vs. 30.6%; p = 0.570) and uniformly high continence rates (ICIQ-UI ≤ 6: 100.0% vs. 96.8%; p = 0.496) in the later phase, though neither comparison reached statistical significance. RARP demonstrates a measurable early institutional learning curve. Using CUSUM analysis, progressive improvement in oncological outcomes, particularly in non-organ-confined disease, was observed with increasing case volume. These findings provide pragmatic benchmarks for emerging robotic centres, though the learning curve for RARP is multifactorial and extends beyond the scope of this initial series.